Traffic flow breakdown, characterized by the abrupt transition from smooth traffic to congestion, impacts motorways globally, causing delays, fuel wastage, and stress for drivers. Detecting and forecasting these incidents is crucial for motorway safety and efficiency and in aiding Intelligent Transport Systems like automatic incident detection and variable speed limits in mitigating traffic breakdowns. Ireland’s M50 motorway, accommodating around 115,000 daily vehicles, frequently experiences traffic breakdowns. Manual variable speed limits have been implemented to enhance traffic management, and the M50 is equipped with real-time data collection sensors like induction loops and weather sensors. While deep neural networks have gained prominence in predicting traffic breakdowns, their data-intensive nature and longer training times still make artificial neural networks a compelling option due to their simplicity. This paper uses data collected on the M50 motorway in 2019 to train an artificial neural network to forecast traffic breakdowns within the next five minutes on specific M50 sections. Using a genetic algorithm approach, the input parameters for the neural networks were optimized, resulting in a significant improvement from 20% to 62% accuracy over naive parameter selection. Notably, the study reveals that critical features for predicting traffic breakdowns are detectable as early as 30 min before occurrence.